Upload via HF Hub [training_history.json]
Browse files- training_history.json +85 -0
training_history.json
ADDED
|
@@ -0,0 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"train_align_loss": [
|
| 3 |
+
0.087992171685283,
|
| 4 |
+
0.043536608680509604,
|
| 5 |
+
0.03890530138204877
|
| 6 |
+
],
|
| 7 |
+
"train_distill_loss": [
|
| 8 |
+
0.21146803635817307,
|
| 9 |
+
0.0825336162860577,
|
| 10 |
+
0.07123976487379807
|
| 11 |
+
],
|
| 12 |
+
"train_combined_loss": [
|
| 13 |
+
0.29946020818673647,
|
| 14 |
+
0.1260702243934457,
|
| 15 |
+
0.11014506614838655
|
| 16 |
+
],
|
| 17 |
+
"val_align_loss": [
|
| 18 |
+
0.05304378426323334,
|
| 19 |
+
0.04008272755891085,
|
| 20 |
+
0.03956587705761194
|
| 21 |
+
],
|
| 22 |
+
"val_distill_loss": [
|
| 23 |
+
0.10595703125,
|
| 24 |
+
0.08424886067708333,
|
| 25 |
+
0.0848388671875
|
| 26 |
+
],
|
| 27 |
+
"val_combined_loss": [
|
| 28 |
+
0.15900081551323333,
|
| 29 |
+
0.12433158823599418,
|
| 30 |
+
0.12440474424511194
|
| 31 |
+
],
|
| 32 |
+
"val_cosine_per_layer": [
|
| 33 |
+
{
|
| 34 |
+
"18": 0.8968261480331421,
|
| 35 |
+
"19": 0.9204645752906799,
|
| 36 |
+
"20": 0.9347640872001648,
|
| 37 |
+
"21": 0.943111777305603,
|
| 38 |
+
"22": 0.9506996273994446,
|
| 39 |
+
"23": 0.9528549909591675,
|
| 40 |
+
"24": 0.9550282955169678,
|
| 41 |
+
"25": 0.9570911526679993,
|
| 42 |
+
"26": 0.9576565027236938,
|
| 43 |
+
"27": 0.9603378176689148,
|
| 44 |
+
"28": 0.9638790488243103,
|
| 45 |
+
"29": 0.9631733894348145,
|
| 46 |
+
"30": 0.9614765048027039,
|
| 47 |
+
"31": 0.9568784236907959,
|
| 48 |
+
"32": 0.9302677512168884
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"18": 0.912168562412262,
|
| 52 |
+
"19": 0.9382794499397278,
|
| 53 |
+
"20": 0.9520896673202515,
|
| 54 |
+
"21": 0.9596585631370544,
|
| 55 |
+
"22": 0.9656936526298523,
|
| 56 |
+
"23": 0.9666958451271057,
|
| 57 |
+
"24": 0.9676884412765503,
|
| 58 |
+
"25": 0.9690631628036499,
|
| 59 |
+
"26": 0.9692828059196472,
|
| 60 |
+
"27": 0.970467746257782,
|
| 61 |
+
"28": 0.9725496768951416,
|
| 62 |
+
"29": 0.9717246890068054,
|
| 63 |
+
"30": 0.9690486788749695,
|
| 64 |
+
"31": 0.9655911922454834,
|
| 65 |
+
"32": 0.9485434889793396
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"18": 0.9129157662391663,
|
| 69 |
+
"19": 0.9390837550163269,
|
| 70 |
+
"20": 0.9528340697288513,
|
| 71 |
+
"21": 0.960339367389679,
|
| 72 |
+
"22": 0.9662702083587646,
|
| 73 |
+
"23": 0.96719890832901,
|
| 74 |
+
"24": 0.9681627154350281,
|
| 75 |
+
"25": 0.9694998264312744,
|
| 76 |
+
"26": 0.9696952700614929,
|
| 77 |
+
"27": 0.9708241820335388,
|
| 78 |
+
"28": 0.9728708863258362,
|
| 79 |
+
"29": 0.9720371961593628,
|
| 80 |
+
"30": 0.9692854881286621,
|
| 81 |
+
"31": 0.9659169316291809,
|
| 82 |
+
"32": 0.9493541121482849
|
| 83 |
+
}
|
| 84 |
+
]
|
| 85 |
+
}
|